Resources
ICH M7 in practice: classifying and controlling mutagenic impurities
How Class 1 to 5 assignment, the 1.5 microgram per day threshold and the four control options change what you need from a reference standard.
Published 26 September 2026 · ImpurityLab technical team
What M7 actually asks
ICH M7 asks two questions in sequence, and answers them in a fixed order. First, is this impurity DNA-reactive, that is mutagenic? Second, if it is, how much of it is acceptable?
The order matters because the first question is answered structurally and computationally before it is answered experimentally. Two complementary in-silico methodologies — an expert rule-based system and a statistical system — are used to look for structural alerts. If both predict negative, the guidance allows the compound to be treated as non-mutagenic without an Ames test. If either predicts positive, an Ames test can be used, and a negative result actually overrides the prediction.
The practical consequence is that the impurity list gets shorter but far sharper. Fewer compounds matter. For the ones that do, the acceptable limit is usually one or two orders of magnitude below the reporting threshold you already work to under ICH Q3A.
M7 applies to DNA-reactive impurities. Everything else stays under the Q3A and Q3B framework with their percentage-based reporting, identification and qualification thresholds. The two systems run in parallel on the same drug substance, and an impurity can be in scope for both.
The five classes
| Class | Definition | What it changes |
|---|---|---|
| Class 1 | Known mutagenic carcinogens — bacterial mutagenicity positive and carcinogenic in animals | Compound-specific acceptable intake derived from carcinogenicity data. No default applies. |
| Class 2 | Known mutagens with unknown carcinogenic potential — bacterial mutagenicity positive, no useful carcinogenicity data | Default acceptable intake of 1.5 µg per day. |
| Class 3 | Alerting structure unrelated to the drug substance, with no mutagenicity data | Treated as Class 2 until data exist. An Ames test can move it to Class 5. |
| Class 4 | Alerting structure shared with the drug substance, or the compound is known to be non-mutagenic but relevant data are incomplete | Needs experimental data to resolve. A negative Ames test moves it to Class 5; a positive result makes it Class 2. |
| Class 5 | No structural alert, or an alert with sufficient data demonstrating the compound is non-mutagenic | No mutagenic control required. Managed as an ordinary impurity under Q3A. |
Note where most of the work lands. Class 3 and Class 4 are unresolved states, not conclusions. They exist so that the classification can be made early from structure, then closed later with data. Once a bacterial mutagenicity result exists, almost every compound ends up in Class 1, 2 or 5.
The threshold that drives the analysis
For Class 2 and Class 3 impurities the default acceptable intake is a threshold of toxicological concern of 1.5 µg per day. It is not a detection limit and it is not a specification; it is an intake, expressed per day, and it has to be converted into a concentration in your material before a method can be built around it.
The conversion is the same arithmetic that governs nitrosamine methods, and it produces the same uncomfortable numbers:
- Take the acceptable intake — 1.5 µg per day.
- Divide by the maximum daily dose of the product.
- The result is the specification in parts per million.
- Set the working target below that, typically at thirty percent, and derive the required limit of quantitation from the sample concentration.
A high-dose product with a two gram maximum daily dose gives a specification of roughly 0.75 ppm. A low-dose product with a one milligram daily dose gives 1,500 ppm, where the impurities present at that level are unlikely to be the analytical challenge at all. The analytical burden of M7 is concentrated almost entirely in high-dose products.
The threshold is a cumulative concept as well as an individual one. Several Class 2 impurities each present at 0.9 µg per day are not automatically acceptable just because none of them individually exceeds 1.5 µg per day, and the sum across the class is what a reviewer will look at.
Cohort of concern
A small group of structural classes is excluded from the default threshold, on the basis that their potency is not distributed in a way that supports a single default. The classical members of this group are N-nitroso compounds, aflatoxin-like structures and alkyl-azoxy compounds.
For these, a compound-specific acceptable intake derived from carcinogenicity data is the normal route. Where that is not possible, the recent revision of M7 recognises a conservative default for N-nitroso compounds and a structured potency categorisation approach that assigns a value from the available data rather than falling back to a single number for the whole class.
The practical reading is that nitrosamines are not simply “Class 2 with a different number”. They are assessed on their own terms, the assigned value is compound-specific, and it is revised when the potency database changes. Any nitrosamine limit quoted in an old method should be checked against the currently assigned value before the method is used again.
The four control options
M7 does not require that every mutagenic impurity be tested in the drug substance. It offers four control strategies, and the choice is a manufacturing argument rather than an analytical one.
- Test the drug substance against the acceptable limit. The most direct option, and the most expensive in routine testing.
- Test an intermediate at a level that, given the demonstrated purging power of the steps that follow, delivers a drug substance at or below the limit.
- Test upstream at a less stringent limit — at a starting material or early intermediate — taking credit for purge, with the drug substance level predicted rather than measured.
- Do not test at all, and instead justify the conclusion from a scientific assessment of the process: the impurity is not formed, or the process removes it with a demonstrated margin.
Options three and four both rest on a purge argument, and a purge argument rests on a number. The predicted purge factor — the ratio of the amount entering a step to the amount leaving it, multiplied across the relevant unit operations — has to exceed the required purge factor, which is the ratio between the level that would otherwise be present and the acceptable limit. Where the required purge factor is large — and for a potent Class 1 impurity it can be very large — the argument becomes demanding, and spiking studies are usually needed to support it.
What this changes for your reference standard
Three things, and all three are procurement decisions rather than analytical ones.
- The limit is low enough that the standard's own impurity content matters. When the specification is a fraction of a part per million, an impurity present at one percent in the reference material consumes a measurable share of the signal you are using to control it. The chromatogram becomes part of the specification.
- Identity has to be structural, not nominal. Class assignment depends on the structure. A standard whose certificate gives a name and a CAS number but no structural confirmation cannot support the assignment it is being used to justify.
- Isotope-labelled analogues become necessary rather than convenient. At these concentrations the method is quantitative only if losses and matrix effects are corrected. That means a labelled analogue of the specific impurity, not a generic internal standard.
None of the three is exotic. They are all things a supplier either builds in or does not, and the difference shows up as time in the method development rather than as cost in the catalogue.
The short version
Classify from structure, close the classification with data, convert the acceptable intake into a specification using the maximum daily dose, and then decide whether you are testing the drug substance or arguing that you do not need to. Nitrosamines sit outside the default and are assigned their own numbers. Everything else stays where it was, under a percentage-based threshold that is far more forgiving than people expect.